This week, as DeepSeek, a free AI-powered chatbot from China, embarrassed American tech giants and panicked investors, sending global markets tumbling, investor Marc Andreessen described its emergence as “AI’s Sputnik moment.” That is, the moment when self-belief and confidence tips over into hubris. It was not just stock prices that plummeted. The carefully constructed story of American technological supremacy also took a deep plunge.
But perhaps the real shock should be that Silicon Valley was shocked at all.
For years, Silicon Valley and its cheerleaders spread the narrative of inevitable American dominance of the artificial intelligence industry. From the “Why China Can’t Innovate” cover story in the Harvard Business Review to the breathless reporting on billion-dollar investments in AI, U.S. media spent years building an image of insurmountable Western technological superiority. Even this week, when Wired reported on the “shock, awe, and questions” DeepSeek had sparked, the persistent subtext seemed to be that technological efficiency from unexpected quarters was somehow fundamentally illegitimate.
“In the West, our sense of exceptionalism is truly our greatest weakness,” says data analyst Christopher Wylie, author of MindF*ck, who famously blew the whistle on Cambridge Analytica in 2017.
That arrogance was on full display just last year when OpenAI’s Sam Altman, speaking to an audience in India, declared: “It’s totally hopeless to compete with us. You can try and it’s your job to try but I believe it is hopeless.” He was dismissing the possibility that teams outside Silicon Valley could build substantial AI systems with limited resources.
There are still questions over whether DeepSeek had access to more computing power than it is admitting. Scale AI chief executive Alexandr Wong said in a recent interview that the Chinese company had access to thousands more of the highest grade chips than people know about, despite U.S. export controls. What’s clear, though, is that Altman didn’t anticipate that a competitor would simply refuse to play by the rules he was trying to set and would instead reimagine the game itself.
By developing an AI model that matches—and in many ways surpasses—American equivalents, DeepSeek challenged the Silicon Valley story that technological innovation demands massive resources and minimal oversight. While companies like OpenAI have poured hundreds of billions into massive data centers—with the Stargate project alone pledging an “initial investment” of $100 billion—DeepSeek demonstrated a fundamentally different path to innovation.
“For the first time in public, they’ve provided an efficient way to train reasoning models,” explains Thomas Cao, professor of technology policy at Tufts University. “The technical detail is that they’ve come up with a way to do reinforcement learning without supervision. You don’t have to hand-label a lot of data. That makes training much more efficient.”
By developing an AI model that matches—and in many ways surpasses—American equivalents, DeepSeek challenged the Silicon Valley story that technological innovation demands massive resources and minimal oversight.
For the American media, which has drunk the Silicon Valley Kool Aid, the DeepSeek story is a hard one to stomach. For a long time, Wylie argues, while countries in Asia made massive technological breakthroughs, the story commonly told to the American people focused on American tech exceptionalism.
An alternative approach, Wylie says, would be to see and “acknowledge that China is doing good things we can learn from without meaning that we have to adopt their system. Things can exist in parallel.” But instead, he adds, the mainstream media followed the politicians down the rabbit hole of focusing on the “China threat.”
These geopolitical fears have helped Big Tech shield itself from genuine competition and regulatory scrutiny. The narrative of a Cold War style “AI race” with China has also fed the assumption that a major technological power can be bullied into submission through trade restrictions.
That assumption has also crumpled. The U.S. has spent the past two years attempting to curtail China’s AI development through increasingly strict controls on advanced semiconductors. These restrictions, which began under Biden in 2022 and were significantly expanded last week under Trump, were designed to prevent Chinese companies from accessing the most advanced chips needed for AI development.
DeepSeek developed its model using older generation chips stockpiled before the restrictions took effect, and its breakthrough has been held up as an example of genuine, bootstrap innovation. But Professor Cao cautions against reading too much into how export controls have catalysed development and innovation at DeepSeek. “If there had been no export control requirements,” he said, “DeepSeek could have been able to do things even more efficiently and faster. We don’t see the counterfactual.”
DeepSeek is a direct rebuke to both Western assumptions about Chinese innovation and the methods the West has used to curtail it.
As millions of Americans downloaded DeepSeek, making it the most downloaded app in the U.S., OpenAI’s Steven Heidel peevishly claimed that using it would mean giving away data to the Chinese Communist Party. Lawmakers too have warned about national security risks and dozens of stories like this one echoed suggestions that the app could be sending U.S. data to China.
Security concers aside, what really sets DeepSeek apart from its Western counterparts is not just efficiency of the model, but also the fact that it is open source. Which, counter-intuitively, makes a Beijing-funded app more democratic than its Silicon Valley predecessors.
In the heated discourse surrounding technological innovation, “open source” has become more than just a technical term—it’s a philosophy of transparency. Unlike proprietary models where code is a closely guarded corporate secret, open source invites global scrutiny and collective improvement.
DeepSeek is a direct rebuke to Western assumptions about Chinese innovation and the methods the West has used to curtail it.
At its core, open source means that the source code of a software is made freely available for anyone to view, modify, and distribute. When a technology is open source, users can download the entire code, run it on their own servers, and verify every line of its functionality. For consumers and technologists alike, open source means the ability to understand, modify, and improve technology without asking permission. It’s a model that prioritizes collective advancement over corporate control. Already, for instance, the Chinese tech behemoth Alibaba has released a new version of its own large language model that it says is an upgrade on DeepSpeak.
Unlike ChatGPT or any other Western AI system, DeepSource can be run locally without giving away any data. “Despite the media fear-mongering, the irony is DeepSeek is now open source and could be implemented in a far more privacy-preserving way than anything offered by Meta or OpenAI,” Wylie says. “If Sam Altman open sourced OpenAI, we wouldn’t look at it with the same skepticism, he would be nominated for the Nobel Peace Prize.”
The open-source nature of DeepSeek is a huge part of the disruption it has caused. It challenges Silicon Valley’s entire proprietary model and challenges our collective assumptions about both AI development and global competition. Not surprisingly, part of Silicon Valley’s response has been to complain that Chinese companies are using American companies’ intellectual property, even as their own large language models have been built by consuming vast amounts of information without permission.
This counterintuitive strategy of openness coming from an authoritarian state also gives China a massive soft power win that it will translate into geopolitical brownie points. Just as TikTok’s algorithms outmaneuvered Instagram and YouTube by focusing on accessibility over profit, DeepSeek, which is currently topping iPhone downloads, represents another moment where what’s better for users—open-source, efficient, privacy-preserving—challenges what’s better for the boardroom.
We are yet to see how DeepSeek will reroute the development of AI, but just as the original Sputnik moment galvanized American scientific innovation during the Cold War, DeepSeek could shake Silicon Valley out of its complacency. For Professor Cao the immediate lesson is that the US must reinvest in fundamental research or risk falling behind. For Wylie, the takeaway of the DeepSeek fallout in the US is more meta: There is no need for a new Cold War, he argues. “There will only be an AI war if we decide to have one.”
Additional reporting by Masho Lomashvili.